Skip to content

Batch scheduling and changeover minimisation

Every changeover is lost production time. The order of your batches decides how much you lose. Our model finds the order that loses the least.

  • Free 30-minute call
  • No commitment
  • You own the code
Typical industries
Manufacturing, Pharma and life sciences
Problem type
Scheduling
You receive
A documented model, run on your data

In short: A changeover-minimising schedule sequences production batches so that the time lost to cleaning, retooling and setup between them is as small as possible. It pays off wherever setup time depends on which product ran before.

The problem

Friday afternoon on the stamping line. The next order is brackets from thick sheet, then small clips from thin sheet. After that come more brackets in another variant. In that order, your crew fits and removes the heavy tool twice. Swap the last two orders, and one tool change is enough. Whether they get swapped depends on who’s on shift.

A changeover is time a machine spends producing nothing. Every switch to a new part, material, grade or size means your crew changes the tool, resets the machine or cleans it. For that stretch, the machine costs money without making parts. The changeover itself is often unavoidable. But you can influence how much changeover time you pay for.

The cost of a switch depends on which two batches you switch between. If the tool stays and only the sheet thickness changes, a quick adjustment is often enough. If the tool changes, the stamping line stops for much longer. And switching back from B to A doesn’t always take as long as switching from A to B. So the order of your batches decides how many machine-hours go into changeovers instead of parts.

Thin to thick, small to large

Every plant already has rules of thumb for this. In extrusion of profiles and pipes, you run thin to thick, or group by die and material grade. The die is the shaped tool the material is pressed through. In machining, you group parts that need the same fixtures and tools. The wrong order forces longer transitions, more scrap while the extrusion line ramps up and sometimes a full die change.

These are sequencing problems by size, material and tool. “Always go thin to thick” captures the idea, but rarely the best order. The rule of thumb ignores due dates and batch volumes. And it misses the times when breaking the rule once saves three changeovers later. If you want production changeover reduction beyond the rule of thumb, you need a model that does the arithmetic.

Signs your changeovers are costing too much

Your changeover hours creep up, even though volumes stay flat. Each shift runs its own sequence. Rush orders force extra tool changes and cleans that weren’t in the plan. You can’t say how many machine-hours a better sequence would free up. Each of these points to capacity you already own but can’t use.

Why generic scheduling software falls short

Generic scheduling software applies a fixed policy: thin to thick, earliest due date or grouping by product family. It has no objective function, a score that judges possible sequences by your criteria. So it can’t tell whether the sequence keeps total changeover time low. When a rush order doesn’t fit the rule, the software can’t work out whether one longer changeover now saves three shorter ones later. Each shift makes that call by feel.

The software doesn’t know what only your plant has: which tool change takes an hour, which machine needs a full clean after which material. We build the model for your plant, and each of your switch times becomes part of it. You check the result on your own orders before you decide.

What changes

Today, each shift sets its own sequence, and nobody sees what it costs in changeover time. Our model proposes the week’s order with every switch cost and cleaning rule built in. You see how much changeover time each switch costs. You check the plan, move a rush order if needed and release it.

SMED or sequencing first?

If you’ve done lean work, you know SMED, single-minute exchange of die. Shigeo Shingo developed it in the 1950s and 1960s, much of it at Toyota. You split a changeover into internal work, done while the machine is stopped, and external work, done while it still runs. Then you turn as much internal work as you can into external work. The target is under ten minutes per changeover.

SMED makes each switch shorter. Sequencing decides which switches you run at all, and the two add up. Even Toyota scheduled its die changes in a standard sequence. We think sequencing is the cheaper first step, because it needs no new tooling and no retraining. A changeover you never run doesn’t need to be made faster.

Batch scheduling optimisation: what you get

Setup time, cleaning frequency and due dates become goals the model balances. Your switch costs, cleaning rules and campaign lengths become hard constraints it never violates. A campaign is a block of similar batches you run in one go, with no big switch inside. We test the model with your orders and in your weeks with the most rush orders before it goes live. You judge it on setup hours, changeover count, on-time delivery and machine utilisation.

  • A model written around your own batch scheduling and changeover minimisation rules.
  • Results measured in the numbers you already track.
  • It slots into how you already work, and you can check every answer it gives.

How it works

Every shift, job and resource goes into one model, and your rules go in as hard limits. You get the best schedule that keeps all of them. What “best” means is your call: lowest cost, the fairest split or the fewest late orders.

Sequence-dependent setup times (SDST)

When setup time depends on the order of the batches, planners call it sequence-dependent setup times, or SDST. Setup time is the time a machine needs to switch from one batch to the next. In an ordinary schedule, it’s a fixed number. With SDST, it’s a matrix, a table with a cost for every pair of batches. Switching from batch A to B can take far more or far less time than switching from B to A. Take a paint plant: white to black takes 15 minutes, because a trace of white disappears in black. Black to white takes two hours, because the mixing tank needs a full clean. Black to dark grey takes only 20 minutes again.

That asymmetry makes batch sequencing its own problem. You don’t only decide when each batch runs. You also look for the order with the lowest total setup time. Every switch you pick changes how long the next one takes. An SDST scheduling algorithm is built to search all those possible orders.

Why changeover sequencing is NP-hard

Formally, this is close to the asymmetric travelling salesman problem. A salesman must visit every city once, on the shortest round trip. In the asymmetric version, A to B can cost more than B to A. Here the cities are your batches, and the distances are your changeover times. That problem is NP-hard: no known method finds the proven best answer quickly once it grows.

The number of possible orders grows factorially, which means each extra batch multiplies it. Ten batches can already run in 3.6 million different orders. The useful question is how good a sequence you get in the time you have.

Cleaning runs and campaign planning

Two more things shape the plan: cleaning runs and campaigns. The first is cleaning-run scheduling. In food, chemicals and pharma, certain switches require a full clean. Think of an allergen changeover or a validated wash. A validated wash follows a cleaning procedure that has been tested, proven and documented.

Each clean is a big, fixed cost you want to trigger as rarely as the rules allow. So you group batches that share a cleaning class, meaning products with the same cleaning needs. One clean then covers many batches. The second structure is campaign planning: how long each block of similar product runs before the next changeover.

Longer campaigns save setups, but they build inventory and put due dates at risk. The model sets how long each campaign runs, so you change over as little as possible and still meet your due dates. In pharmaceutical manufacturing, campaign planning and validated cleaning dominate the schedule. That’s why changeover planning pays off in pharma just as much as in general manufacturing.

Which method fits your problem

Where the problem is small enough, we use mixed-integer linear programming (MILP). It models every switch as a yes-or-no decision and proves the best order. Where exact solving would take too long, we use large neighbourhood search, which tears up part of the sequence and rebuilds it better. Or we use tabu search, which bans its own recent moves for a while so it can’t circle back. Neither proves that no better order exists, but both find strong sequences in the time you have.

Common questions.

Something else? Ask us directly.

We’d start with the sequence, though the two add up. SMED stands for Single-Minute Exchange of Die. It shortens each changeover: you do as many steps as you can while the machine is still running. The sequence decides which changeovers you need at all. It needs no new tooling and shows quickly which changeovers cost the most time. Those are then the best targets for a SMED workshop.

Yes. A changeover matrix is a table with the changeover time for each pair of orders. Most plants know their changeovers by product family, even if nobody wrote them down per pair. We start with families and rough times from your supervisors, then refine with logged data. A matrix that’s roughly right already lets the model compare sequences. Precision matters most for the few switches that cost the most.

They can, and that’s the trade-off the model makes visible. A campaign is a run of similar orders that you make one after another. Longer campaigns mean fewer changeovers, but more stock, and due dates are more likely to slip. With shorter campaigns you deliver more reliably but lose more time on changeovers. The model sets the campaign length for each product, and you set what a day of stock costs you.

Sometimes it can, and if that’s the case, we’ll tell you. Standard software is built for an average operation, though. Yours runs on its own rules: your contracts, your exceptions, the things you know by heart. A custom model writes those rules in, and that’s where the better plan comes from.

You don’t need either to start. We build the model from the data you’ve already got, even if it lives in three spreadsheets and one person’s head. You get a system you run yourself. We train your team until they no longer need us.

You see a first result with the prototype, before you commit to anything bigger. We start with a free 30-minute call about your operation and the model that would fit it. Then we build a working prototype and run it on your numbers.

Tell us what you’re planning.

In a free 30-minute call with a specialist, you find out whether optimisation can help.

  • Free, no commitment
  • You talk to a specialist directly
  • A clear go or no-go